Terence Tao: AI is 'non-renewably mining' open math problems

Fields Medalist Terence Tao has warned that the indiscriminate use of AI to solve open mathematical problems may be "non-renewably mining" the field's intellectual resources. Tao argues that while these tools can achieve immediate results, they risk undermining the long-term ecosystem required for future mathematical progress.

Writing on Mathstodon, Tao expressed concern that powerful solution-extraction tools are being used to clear away open problems without the accompanying human analysis that typically drives the field forward. He noted that the indiscriminate use of such tools can solve the problems at hand, but does so "at the cost of sustaining the ecosystem for the next wave of progress."

The Value of the Struggle

This warning comes as the discipline shifts toward "Big Mathematics," a hybrid approach combining human insight, AI generation, and formal verification systems like Lean. As models from organizations such as DeepMind and OpenAI increasingly tackle complex conjectures, the process of discovery is changing. Traditionally, the value of a mathematical breakthrough lies not just in the final proof, but in the new intuitions and conceptual tools developed during the struggle to find that proof.

Tao argues that raw AI solutions, when stripped of careful human analysis, can be of "negligible or even negative value" for understanding the difficulty landscape of nearby problems. In this view, a solution provided by an AI without a clear, human-understandable path of reasoning fails to provide the insights necessary for mathematicians to tackle subsequent, more difficult challenges.

Implications for Creativity

If AI is used to harvest the "low-hanging fruit" of open problems without humans understanding the underlying logic, the mathematical community may face a paradox: an increase in the number of solved theorems alongside a stagnation in human mathematical creativity. The loss of these conceptual stepping stones could leave future researchers without the intuition needed to formulate new theories or approach unsolved mysteries.

A Path Toward Sustainable Discovery

To mitigate this risk, Tao advocates for a more disciplined approach to AI integration. He suggests designating certain classes of problems as requiring careful analysis, ensuring that the goal remains the identification of insights from the solution process rather than mere raw extraction.

What remains to be seen is whether the mathematical community will adopt formal guidelines to protect these "intellectual reserves" or if the acceleration of AI capabilities will outpace the human ability to analyze them. For now, Tao's warning serves as a call to prioritize the process of discovery over the mere accumulation of answers.